1732 posts tagged with "Self-Hosting"
Running your own PaaS and infrastructure on machines you own
From Azure to Hetzner: What a 2026 Switchers' Guide Gets Right About the Hyperscaler Exit Wave
A 2026 Azure-to-Hetzner switchers' guide promises 70–90% savings; this teardown verifies the egress math and unbundling map, prices the re-assumed ops labor, and distills a PaaS migrator's runbook.
Your Fleet Runs on Go 1.26, Not Go 1.27: The Kubernetes 1.37 Toolchain-Coupling Checklist
Kubernetes 1.37 ships built on Go 1.26, not Go 1.27, so a fleet that builds its own controllers, providers, and node images inherits stdlib CVEs on two different schedules. Here is the version truth table and the ordered rebuild checklist.
Helm 3 Bug Fixes Are Over: A Fleet Owner's Migration Checklist for the Security-Only Runway to February 2027
Helm 3's last bug-fix release shipped September 9, 2026, and security patches end February 10, 2027, with no extended support to buy. This fleet-owner's checklist covers what 'security fixes only' actually means, how to choose between Helm 4, plain manifests, and a GitOps operator, and the four automation breakages to fix before the wall.
Heroku Put Cloud Native Buildpacks on Legacy Cedar Apps: Why OCI-Via-Buildpacks Is Now the Safe Build Bet
Heroku's September 3, 2026 changelog brought Cloud Native Buildpacks to Cedar-generation apps — what the four build-artifact deltas change, why a sustaining-mode platform spent engineering on convergence, and the default-to-CNB build strategy for git-push platforms.
Hetzner's 2026 Price Hikes and the 20TB-to-0.5TB Traffic Trap: Recomputing Self-Hosted Fleet Unit Economics
Hetzner raised cloud prices twice in 75 days in 2026 — the CCX13 jumped 169% to €42.99 while the ARM CAX11 held at €5.99 — and traffic allowances collapse from 20TB in the EU to 1TB in the US and 0.5TB in Singapore with €7.40/TB overage. This teardown recomputes per-vCPU, per-GB, and per-TB unit costs across the compute ladder, Server Auction metal, and all three regions, plus five rules for where just-add-another-box still works.
Hetzner Gives Away the Model Calls: What Free Inference Is Actually Worth to an Agent Builder
Hetzner's free experimental Inference API serves Qwen open-weight models at zero per-token cost — a worked comparison shows it replaces about $30 a month of hosted inference for a typical agent workload, within rate limits and no-SLA terms that rule out production use.
Still on ingress-nginx? Your Migration Starts With an Audit, Not a Tool
The ingress-nginx repository has been read-only since March 2026. A close reading of the January retirement statement plus the ordered pre-migration audit — annotation inventory, five routing-behavior probes, default-backend and TLS semantics — to run before any conversion tool.
Half of Kubernetes Got a Two-Month Warning: What the Ingress-NGINX Governance Crisis Means for Self-Hosted Platforms
The Kubernetes Steering and Security Response Committees' joint ingress-nginx warning was a governance signal, not a CVE — here's the escalation ladder, a default-routing decision matrix, and a five-check policy for request-path dependencies.
Kool.dev's $49 BYOC vs. Owning the Whole Stack: A Cost-and-Control Teardown
Kool.dev's BYOC plan runs k0s on your VPS for $49/mo, but the $5 managed databases and $0.45/GB storage reintroduce the vendor dependency self-hosting was meant to remove. A three-column cost teardown with trust-boundary analysis.